Cycle spinning, a technique mainly used for wavelet denoising, has also been shown to be successful toward image resolution upscaling in the wavelet domain. We propose a directional variant of the cycle spinning methodology. We obtain estimates of local edge orientation from a wavelet decomposition of the available low-resolution image and use this information to influence the choice of cycle spinning parameters that are employed for resolution upscaling. Our experimental results show that the proposed method outperforms competing methods for a wide range of images offering modest but consistent improvements both in objective as well as subjective terms. Lower computational complexity compared to the conventional cycle spinning is also demonstrated.

A new multiscale corner detection method is proposed based on dyadic wavelet transform (WT) of the orientation function of a contour image. As the decomposition of the dyadic WT is complete and its scales are sparse, all the scales are defined as natural scales for corner detection. The points that are wavelet transform modulus maxima (WTMM) at different scales are taken as corner candidates. For each corner candidate, the sum of the corresponding normalized WTMM at all the natural scales is used as significance measure of the "cornerness". The utilization of the complete information makes the performance of the proposed detector independent to the type of input images. The decomposition scales of the WT are restricted by the contour length, which makes the algorithm adaptable for both long contours and short contours. Both subjective and objective evaluation illustrate better performance of the proposed corner detector compared to the conventional methods.

In image classification, the common texture-based methods are based on image gray levels. However, the use of color information improves the classification accuracy of the colored textures. In this paper, we extract texture features from the natural rock images that are used in bedrock investigations. A Gaussian bandpass filtering is applied to the color channels of the images in RGB and HSI color spaces using different scales. The obtained feature vectors are low dimensional, which make the methods computationally effective. The results show that using combinations of different color channels, the classification accuracy can be significantly improved.

We deal with the applications of a class of nonlinear dynamic systems to image transformation and encoding whereby the nonlinear system presents a chaotic or hyperchaotic attractor. Several ways are possible to encode or transform the image using chaos. We develop algorithms for image encoding based on the permutation of the pixel value, position, or both. This approach enables the fast decorrelation of relations among pixels in the initial image in a random-like fashion. We illustrate the use of 1-D, 2-D, and 3-D maps for this purpose. We also use chaotic dynamical systems with a single or two outputs. A discussion of the sensitivity of the algorithms to the keys is followed by the illustration of the algorithms using example images.

JPEG2000 Standard (ISO/IEC 15444, ITU-T T.800) is a new-generation image compression technique, enabling encoding images at low bit rates with acceptable quality. Since JPEG2000 is based on wavelet transforms, the reconstructed image will contain perceivable ringing artifacts in medium- and low-bit-rate regimes of lossy compression. We utilize a quad-tree partitioning scheme for postprocessing the reconstructed image in a spatially varying manner and presents a voting strategy to determine a set of morphological filters to be used for reducing the ringing artifacts. All this processing is performed at the encoder side, and the set of selected filters is conveyed to the decoder in the form of side information. Specifically, the adopted (eight) morphological filters are generated through the use of four predefined structuring elements (SEs) in conjunction with two morphological operations, namely, dilation and erosion. A voting strategy is used to select the morphological filter for each partition block to optimize the postprocessed image quality—through minimizing the sum of absolute differences under the constraint of a fixed quad-tree partition. Applying the chosen morphological filters to their corresponding blocks typically enhances the reconstruction quality—in the worst case, leaving it unchanged. Simulation results demonstrate that the proposed technique enhances reconstructed image quality compared to unprocessed JPEG2000 output at an equivalent bit rate, accounting for the side information overhead, in terms of both peak SNR (PSNR) and visible ringing measure (VRM).

The discrete cosine transform (DCT) is widely used in many practical image/video compression systems because of its compression performance and computational efficiency. We adopt the DCT and the modified set partitioning in hierachical trees (SPIHT) algorithm that was designed initially for encoding the discrete wavelet transform (DWT) coefficients to be suitable to encode DCT coefficients. The algorithm represents the DCT coefficients to concentrate signal energy and proposes a combination and dictator to eliminate the correlation in the same level subband for encoding DCT-based images. To further save bits, subbands with significant coefficients are classified into seven types. The coding complexity of the proposed algorithm for DCT coefficients is just close to JPEG but the performance is higher than JPEG2000. Experimental results indicate that the proposed technique improves the quality of the reconstructed image in terms of peak SNR (PSNR) over SPIHT and JPEG2000 at the same bit rate.

Protection of digital multimedia content has become an increasingly important issue for content owners and service providers. As watermarking is identified as a major technology to achieve copyright protection, the relevant literature includes several distinct approaches for embedding data into a multimedia element. Because of its growing popularity, the discrete wavelet transform (DWT) is commonly used in recent watermarking schemes. In a DWT-based scheme, the DWT coefficients are modified with the data that represents the watermark. We present a hybrid nonblind scheme based on DWT and singular value decomposition (SVD). After decomposing the cover image into four bands (LL, HL, LH, and HH, where L stands for lowpass and H stands for highpass), we apply the SVD to each band, and modify the singular values of the cover image with the singular values of the visual watermark. Modification in all frequencies enables the development of a watermarking scheme that is robust to a wide range of attacks. We compare our hybrid algorithm with a pure SVD-based scheme, and show that it is considerably more robust and reliable.

We develop a digital camera model with the purpose of simulating the performance of a sky radiance colorimeter, based on a single-sensor camera equipped with a charge-coupled device imager. To provide a color image, the sensor is covered with a Bayer color filter array (CFA). The model input is the radiance, expressed in radiometric units. Demosaicking of the CFA is performed using three common interpolation methods. Color characterization is performed by calculating 3×3 matrices, based on either the maximum-ignorance assumption or computer-generated sky radiance training sets. For the model to be reliable, we incorporate multiple noise sources. We also parameterize many "critical" characteristics of the simulated imager, for example, temperature-related dark-current generation and pixel size. Error propagation analysis is performed, including an analysis for the demosaicking methods. Colorimetric performance results, expected measurement accuracy, and validation of the derived noise propagation equations are presented.

A novel image enhancement algorithm called AINDANE (adaptive and integrated neighborhood dependent approach for nonlinear enhancement) for improving the visual quality of digital images captured under extremely low or nonuniform lighting conditions is proposed. AINDANE is comprised of two separate processes, namely, adaptive luminance enhancement and adaptive contrast enhancement, to provide more flexibility and better control over image enhancement. Adaptive luminance enhancement is a global intensity transformation based on a specifically designed nonlinear transfer function, which is self-tuned by the histogram statistics of the input image. This process largely increases the luminance of darker pixels and compresses the dynamic range of the image at the same time. Adaptive contrast enhancement tunes the intensity of each pixel based on its relative magnitude with respect to the neighboring pixels. This process is also adaptively controlled by the global statistics of the image. A color restoration process, based on the relationship between the spectral bands and the luminance of the original image, is applied to convert the enhanced intensity image back to a color image.

The Kullback-Leibler (KL) divergence, which is a fundamental concept in information theory used to quantify probability density differences, is employed in assessing the color content of digital images. For this purpose, digital images are encoded in the CIELAB color space and modeled as discrete random fields, which are assumed to be described sufficiently by 3-D probability density functions. Subsequently, using the KL divergence, a global quality assessment of an image is presented as the information content of the CIELAB encoding of the image relative to channel capacity. This is expressed by an image with "maximum realizable color information" (MRCI), which we define. Additionally, 1-D estimates of the marginal distributions in luminance, chroma, and hue are explored, and the proposed quality assessment is examined relative to KL divergences based on these distributions. The proposed measure is tested using various color images, pseudocolor representations and different renderings of the same scene. Test images and a MATLAB implementation of the measure are available online at http://www.ellab.physics.upatras.gr/PersonalPages/VTsagaris/research.htm.

We propose a new algorithm to digitally restore vintage photographic prints affected by foxing and water blotches. It semiautomatically recovers the defects utilizing the features of the stains. The restoration process enhances the residual image information still present in the area. It is composed of three different steps: inpainting, additive-multiplicative (A-M) modeling, and interpolation.

We propose the Bayesian image reconstruction with prior edges (BIRPE) algorithm for reconstructing an image from Fourier-domain samples with prior edge information from a higher resolution image. A major difference between BIRPE and previous methods is that all edges are detected automatically, and no segmentation of the prior image is required. Also, an edge found in the prior image does not need to be confirmed by the observations; smoothing is reduced across the edge if either the prior image or the observations suggest an edge. Simulations and results on magnetic resonance spectroscopic data are presented that demonstrate the effectiveness of the BIRPE method.

We present an edge detection method based on a water-flow model and gradient information. The gradient magnitude image emphasizes edges of objects in an image and effective extraction of well-defined and connected edges having large gradient values follows. The proposed method can be classified as a locally adaptive thresholding method. To show the effectiveness of the proposed method, its simulation results for various noise-free and noisy synthetic and real images are compared with those of conventional methods. In addition, edge evaluation results of five edge detection methods are shown quantitatively.

We describe a new fully unsupervised image segmentation method based on a Bayesian approach and a Potts-Markov random field (PMRF) model that are performed in the wavelet domain. A Bayesian segmentation model, based on a PMRF in the direct domain, has already been successfully developed and tested. This model performs a fully unsupervised segmentation, on images composed of homogeneous regions, by introducing a hidden Markov model (HMM) for the regions to be classified, and Gaussian distributions for the noise and for the pixels pertaining to each region. The computation of the posterior laws, deduced from these a priori distributions for the pixels, is done by a Markov chain Monte Carlo (MCMC) approach and uses a Gibbs sampling algorithm. The use of a high number of iterations to reach convergence in a segmentation, where the number of segments, or "classes" labels, is important, makes the algorithm rather slow for the processing of a large quantity of data like image sequences. To overcome this problem, we take advantage of the property of the wavelet coefficients, in an orthogonal decomposition, to be modeled by a mixture of two Gaussians. Thus, by projecting an observable noisy image in the wavelet domain, we are able to segment, in this same domain, the wavelet subbands in only two classes. After a decomposition up to a scale J, the main idea is to segment the coarse, and small, approximation subband with a high number of classes, and to segment all the detail (wavelet) subbands with only two classes. The segmented wavelet domain coefficients are then reconstructed to obtain a final segmented image in the direct domain. Our tests on synthetic and natural images show that the segmentation quality stays good, even with noisy images, and shows that the segmentation times can be significantly reduced.

To successfully segment a transmission x-ray image the inherent blurring or structural unsharpness present in such an image must be taken into account. Initial work with an established wavelet-based segmentation algorithm resulted in an oversensitive performance with regard to the number of segmented regions detected. To improve on this performance, two segmentation algorithms are developed and tested. The approaches utilize a wavelet transform together with region merging and, alternatively, multilevel thresholding with region merging. These techniques outperform the ordinary wavelet method. Manually segmented images representing ground truth are used for a comparative evaluation of the segmentation techniques. Our work is part of an ongoing collaborative research program with the U.K. Home Office to analyze information extracted from dual-energy x-ray images for aviation security screening applications.

A fast two-stage scheme for the search of the motion vectors under varying interframe brightness characteristics, referred to as enhanced motion vectors, is devised. In the first stage of the scheme, a given block and the corresponding blocks in the search window are mapped into the sum-of-pixel value domain, where two subsets of candidate blocks, one consisting the blocks having the DC values closest to the DC value of the block of interest and the other consisting of those having the farthest DC values are selected. In the second stage, the motion vector is determined by employing these subsets and using the mean square error as the matching criterion. Experimental results show that the proposed technique provides a high prediction accuracy with a low computational load.

When skin areas such as faces and hands are imaged under natural environments their color appearance is frequently affected by variations in illumination intensity and chromaticity. In color-based skin tracking and detection, changing intensity is often dismissed either with the use of normalized, intensity-invariant color coordinates or by additionally modeling possible skin intensities. Chromaticity variations are rarely considered, although they are common in practice. In most approaches considering chromaticity, the experiments are done with a small or undefined variation range. It is difficult to compare different approaches and assess their applicability range for this reason. To improve the situation, we evaluate the performance of four state-of-the-art methods under drastic but practically common illumination changes. The effect of illumination chromaticity for skin is clearly defined, and based on it we draw conclusion about the performance of these approaches.

We present the work developed on off-line signature verification using hidden Markov models (HMM; Rabiner and Juang, 1986). Left-to-right (LR)-HMM is a well-known technique used by other biometric features, for instance, in speaker recognition and dynamic or on-line signature verification. Our goal here is to extend LR-HMMs to the field of static or off-line signature processing using results provided by image connectivity analysis, which separates images in connected components known as blobs, each one made up of a cluster of adjacent pixels of the same nature. The chain encoding of perimeter points for each blob obtained by this analysis is an ordered set of points in the space, clockwise around the perimeter of the blob. Two different ways of generating models are discussed, depending on the way that blobs provided by the connectivity analysis are ordered. In the first proposed method, blobs are ordered according to their perimeter length. In the second one, blobs are ordered in their natural reading order, i.e., from top to bottom and left to right. Finally, two LR-HMM models are trained using the parameters obtained by these techniques. Verification results corresponding to each technique are compared and some improvements are proposed.

We present a novel technique for extracting bits from the perceptually significant components of an image transformation, thus making the recognition of objects under nonideal conditions robust. Specifically, we describe five popular face recognition transform methods [including principal component analysis (PCA), linear discriminant analysis (LDA), wavelet transform, wavelet transform with PCA, and wavelet transform with Fourier-Mellin transform] with robust bit extraction enhancement for various numbers of bits extracted. The robustness guarantees that all similar face images will produce almost the same bits. This property is useful for generating cryptographic keys. The theoretical results are evaluated on the Olivetti Research Laboratory (ORL) face database, showing that the extended methods significantly outperform the corresponding standard methods when the number of extracted bits reaches 100.

An algorithm for automatic localization of call numbers embedded in book images is proposed. The algorithm first detects potential character pixels by computing the difference and the position relation between the maximum and minimum gradient in a local window. These pixels are combined to form connected components (CCs). Then, CC-based filtering based on the geometry and shape information is performed to eliminate some nontext components, and the retained components are grouped into text lines candidates. Finally, the call number is selected from the text candidates by using a multilayer perceptron (MLP) classifier. Tests are performed on a set of the images acquired under the different conditions, and experimental results prove that the proposed method is effective.

Forensic odontology has long been carried out by forensic experts of law enforcement agencies for postmortem identification. We address the problem of developing an automated system for postmortem identification using dental records (dental radiographs). This automated dental identification system (ADIS) can be used by law enforcement agencies as well as military agencies throughout the United States to locate missing persons using databases of dental x rays of human remains and dental scans of missing or wanted persons. Currently, this search and identification process is carried out manually, which makes it very time-consuming in mass disasters. We propose a novel architecture for ADIS, define the functionality of its components, and describe the techniques used in realizing these components. We also present the performance of each of these components using a database of dental images.

To determine the best match to a query image in a database, conventional content-based image retrieval schemes require the exhaustive search, where the descriptor of the query, e.g., the histogram, must be compared with literally all images in the database. However, the straightforward exhaustive search algorithm is computationally expensive. Thus, fast exhaustive search algorithms are required. We present a fast exhaustive search algorithm based on a multiresolution descriptor structure and a norm-sorted database. First, we derive a condition to eliminate unnecessary matching operations from the search procedure by using a norm-sorted structure of the database. Then, we propose a fast search algorithm based on the elimination condition, which guarantees an exhaustive search for either the best match or multiple best matches to a query. With a luminance histogram as a descriptor, we show that the proposed algorithm provides a search accuracy of 100% with a high search speed.

Some new understanding of a single-lens binocular stereovision system using a biprism (2F filter) is presented. One image captured by this system is divided into two halves and assumed to be one stereo image pair captured by two virtual cameras generated by the biprism. Hence, this system can also be called a virtual stereovision system. Two different approaches to understanding this system are introduced. One approach is based on a camera calibration technique and another is based on a geometrical analysis of ray sketching. Both approaches enable this system to perform depth recovery in a close range like a typical stereovision system. As the approach based on geometrical analysis requires no complex calibration, great implementation and calibration effort can be saved, in contrast with a normal stereovision system. This approach provides a way to build a binocular stereovision system with a simpler implementation but sufficient accuracy. A complete analysis of this system and the related calibration, depth recovery, and experimentation techniques are presented. Experimentation results are presented to prove the effectiveness of both approaches used to understand this system.